Quantification of radioactivity in the body by in vivo bioassay uses counting efficiencies obtained from calibration from a phantom. Usually a standardised BOMAB (Bottle Manikin Absorption) phantom is employed for whole-body counting. The physical size of workers being counted, however, may differ from the calibration phantom, and can be a source of significant errors in dose estimates. A methodology was developed applying subject-specific efficiency data determined by Monte Carlo simulation based on a voxel phantom that was constructed from photographic images of the subject. This approach was demonstrated using a BOMAB phantom. The measured and calculated efficiencies agreed well, with maximum deviation of 30 % at 1.836 MeV (Y-88 gamma-rays). The expected counting efficiencies for an obese volunteer appear higher compared with a BOMAB phantom. This is caused by a closer distance between the detector and the body surface. The fast construction technique of voxel phantoms will contribute to a reduction in uncertainty caused by variations in the counting geometry.